datamllab / tods

TODS: An Automated Time-series Outlier Detection System
http://tods-doc.github.io
Apache License 2.0
1.41k stars 192 forks source link

Maintanance (>=3.8.x, M1, numpy ......) #100

Open nagydavid opened 1 year ago

nagydavid commented 1 year ago

Hi @lhenry15, @yzhao062, @LSC2204 and all contributors.

I am really interested in TODS, I think it is a super good library, however I wonder it is in a pipeline to do some polish on the dependencies, so it can be used in later python releases.

I looked into the recent commits and builds, but after one failed deployment on 3.8 in March, I saw no further attempt to fix the dependency issue.

I am asking is there any chance of keeping this alive or your are going to archive the project.

Best

David

lhenry15 commented 1 year ago

We will keep it alive, based on our recent commit on master branch, it is able to work on Python3.8 (but not w/ M1 chip). Our bottleneck now is the D3M project that served as our backbone is no longer being maintained, and cannot support M1 chip at all (due to the old numpy version). We are still seeking for a good backbone for ML pipeline searching (perhaps Scikit-learn pipeline), but it may need some time for us migrate that. Any idea is appreciated.